1 Introduction
A basic insight from economic theory is that taxation drives a wedge between private and public benefits. This distorts labor supply, consumption, and investment, and leads to loss of welfare. One would therefore expect that in benefit-cost analysis (BCA) of public expenditure, a correction is made for the costs of taxation, i.e., for the marginal excess burden of taxation (MEB) (see, e.g., Pigou, Reference Pigou1928 and references in Section 2).
In line with this, Boardman et al. (Reference Boardman, Greenberg, Vining and Weimer2006) recommend a correction for the MEB. For the USA, it is suggested to use 40 % for federal projects (assuming the income tax is the marginal source of finance) and 17 % for locally financed projects (assuming local real estate tax is the marginal source of finance). The federal BCA guidelines in the USA (OMB, 1992, Circular A94, paragraph 11) recommend an MEB correction of 25 % for public investments.Footnote 2, Footnote 3 A recent article on preschool programs (Heckman et al., Reference Heckman, Moon, Pinto, Savelyev and Yavitz2010) uses three different corrections for the MEB to calculate their social rate of return: 0, 50, and 100 %.
However, in BCA practice all over the world, it is most common not to make a correction. An overview of BCAs on transport infrastructure in Europe (Bickel et al., Reference Bickel, Friedrich, Burgess and Fagiani2006) shows that in only four countries, a correction for the MEB is made. For example, in Denmark and Slovenia, the correction is 20 % and in Sweden, 30 %. In the other countries, no correction is made. In the USA, despite the federal BCA guidelines and the textbook by Boardman et al., no correction is made for BCAs on transport infrastructure,Footnote 4 BCAs on flood control,Footnote 5 BCAs on police interventions on crime,Footnote 6 or the costing of policy measures by the Congressional Budget Office.Footnote 7
The reasons why no correction should be made are often not spelled out or only very briefly discussed. In many BCA guidelines, such as those by the World Bank (1994), the Asian Development Bank (2013), the OECD (2006), and the USA Benefit-Cost Center (Zerbe et al., Reference Zerbe, Davis, Garland and Scott2010), the issue of the MEB is not even mentioned. In the European Union (EU) BCA guidelines for investment in infrastructure financed by the cohesion funds (European Commission, 2014), the issue is hardly discussed, and no correction is recommended unless the national BCA guidelines prescribe this. In the European overview on BCAs on transport infrastructure (Bickel et al., Reference Bickel, Friedrich, Burgess and Fagiani2006, Full report, p. 47), in particular the uncertainty of the estimate of MEB is put forward as a reason not to make a correction.Footnote 8 In some economic literature, it is argued that no correction for MEB is needed in the case of optimal taxation (e.g., Jacobs et al., Reference Jacobs, Mooij and Armstrong2009; Jacobs, Reference Jacobs2018) or when a public good is financed in a distributionally neutral way (e.g., Kaplow, Reference Kaplow1996, Reference Kaplow2004). However, it does not discuss what to do in BCA practice when the theoretical conditions of their model may not apply. In the classic BCA textbook Mishan and Quah (Reference Mishan and Quah2007, p. 240), corrections for the MEB are regarded as “a common error in the many textbooks on the subject.”Footnote 9 It stresses the uncertainty of the MEB estimates, but does not discuss the theoretical perspectives of optimal taxation and distributionally neutral financing.
In the Netherlands, BCA and the correction for MEB has been subject to intense debate among leading economists for some time.Footnote 10 The recently revised Dutch BCA guidelines (Romijn & Renes, Reference Romijn and Renes2013) only state that the issue still has to be clarified. Therefore, the Dutch government has asked a special BCA Working groupFootnote 11 to advise on this issue. The advice should be relevant for all policy areas, not only for transport infrastructure, but also for health care, energy, environmental policies, social security, labor market policy, and tax policy. Broadly in line with the Working group’s report,Footnote 12 this paper investigates the theoretical, empirical, and practical arguments in favor or against an MEB correction.Footnote 13
This paper contributes in various ways to the existing literature. First, with the exception of the book by Dahlby (Reference Dahlby2008), no systematic overview of these arguments exists. Textbooks and guidelines on BCA ignore the issue or provide a limited and unbalanced overview of these arguments, while many academic papers ignore their link with BCA practice. Second, several new arguments are put forward, in particular about the plausibility of the theoretical assumptions of optimal taxation and distributionally neutral taxation. Third, the link between BCA practice and a broad concept of welfare is discussed. This reveals that from a welfare point of view, corrections are needed not only for the MEB, but also for distributional benefits of these taxes and for the distributional benefits of the policy measure itself. Proposals for corrections for the MEB should therefore also be discussed in view of how to take account of these distributional benefits.
Section 2 starts with the perspective from economic theory and explicitly shows the links to the common practice of BCA, i.e., without any correction for MEB. This section includes a discussion not only of the costs of taxation, but also of the distributional benefits of taxation and their link to the practice of BCA and a broad concept of welfare. To this end, a general framework for BCA is presented, in which the costs of taxation and the benefits of redistribution are linked to the other elements that are commonly identified in a BCA.
The framework consists of two parts. The first part starts from the current worldwide BCA practice and covers the costs and benefits without any correction for MEB. The balancing item “net benefits” aggregates costs and benefits irrespective of who gains and who loses and thus follows the principle of “1 euro is 1 euro.” This principle is also known as the Hicks-Kaldor criterion. The second part of the framework shows the corrections for MEB and distributional benefits of taxation and distributional benefits of the policy measure, which are needed to arrive at a broad concept of welfare.
Starting from this framework, Section 3 investigates whether the marginal distributional benefits of taxation are equal to the marginal distortionary effects of taxation. If this is true, then no correction for the MEB is needed and the marginal cost of public funds (MCPF) is equal to 1. This issue will be investigated for financing via the general tax revenues and for other types of financing, e.g., specific types of tax or social security contributions, loans, local taxes, and toll fees.
How to account for distributional benefits of the policy measure in BCA practice is the topic of Section 4. From a welfare perspective, such distributional benefits are important and should be included in the net BCA balance by distributional weighting and not by applying the Hicks-Kaldor criterion, which is most commonly applied in BCA practice all over the world. Conclusions are drawn in Section 5.
2 Costs of taxation, welfare, and benefit-cost analysis
2.1 Marginal cost of public funds and distortionary taxes
The MCPFFootnote 14 can be defined as the ratio of the social value of an extra public euro (i.e., a euro used for a public policy measure) and the social value of an extra private euro (i.e., a euro used for private purposes). Taxation drives a wedge between social and private benefits from economic activities, such as work, entrepreneurship, and schooling. This wedge induces substitution toward less taxed or untaxed activities (e.g., leisure) and this causes a loss of welfare. As a consequence, in order to finance one euro of public expenditure, more than one euro needs to be extracted from the private sector. It is therefore commonly argued that the MCPF is larger than 1 due to the distortionary costs of taxation.Footnote 15
In economic theory, Pigou (Reference Pigou1920)Footnote 16 was the first to advocate that when comparing the benefits and costs of a public good, these distortionary costs of taxation also should be taken into account. In formal economic theory, this idea was taken up by Stiglitz and Dasgupta (Reference Dahlby1971), Diamond and Mirrlees (Reference Diamond and Mirrlees1971), and Atkinson and Stern (Reference Atkinson and Stern1974). They modified Samuelson’s rule on the optimal provision of public goods (Samuelson, Reference Samuelson1954)Footnote 17 to also take tax distortions into account. In the case that public goods are financed by distortionary taxes, this adds to the cost of providing the public goods. This reduces the optimal provision of public goodsFootnote 18 and causes the optimal size of the government to be smaller.
The size of the distortion is different for different taxes. A number of well-known cases are distinguished in economic theory. For example, lump sum taxes do not distort, neither do taxes on goods with inelastic supply or inelastic demand. On the other hand, taxes on wages and investment income distort the supply of labor and decisions on personal saving and investment. Some taxes may even reduce distortion by internalizing negative externalities, e.g., a tax on polluting activities.
2.2 Distributional benefits of taxation
Taxes are needed to finance public expenditure. However, if taxes on wages and investment income are distortionary, this raises the question why, nonetheless, such taxes are in practice preferred to less distortionary taxes. The answer lies in the distributional benefits of such taxes. Taxes with minor distortionary effects,Footnote 19 like fixed levies per inhabitant (“poll tax”), are in particular a heavy burden for households with low income. In order to avoid or limit this, most major taxes are related to income, wealth, or consumption: these are more distortionary, but they reduce inequality.
The distributional benefits of income tax were already stressed by Pigou (Reference Pigou1920, p. 89):
“The old law of diminishing (marginal) utility … leads securely to the proposition: any cause which increases the absolute share of real income in the hands of the poor, provided that it does not lead to a contraction in the size of the national dividend from any point of view, will, in general increase economic welfare.”
This quote also indicates that Pigou was aware that distributional benefits may be accommodated by efficiency losses, i.e., a contraction in the size of the “national dividend”. This trade-off between equity and efficiency is the theme in Okun (Reference Okun1975). His central rule is “Promote equality up to the point where the added benefits of more equality are just matched by the added costs of greater inefficiency.” He argues that a leaky-bucket experiment can test attitudes toward this trade-off. To carry money from the rich to the poor is like transporting water with a leaking bucket, as some money will inevitably disappear. How much leakage will you accept and still support to levy an added tax to the top 5 % of income distribution to benefit the bottom 20 % of income distribution? According to Okun, this requires a judgment on how much the poor need the extra income and how much the rich would be hurt by the extra taxes.
As a consequence, in designing an optimal tax system and in assessing the marginal costs of public funds, distortionary costs of taxation should be regarded as the price to be paid for distributional benefits in terms of reduced inequality. This trade-off between equity and efficiency of the tax system has been further analyzed by Boadway (Reference Boadway1976), Sandmo (Reference Sandmo1998), Slemrod and Yitzhaki (Reference Slemrod and Yitzhaki2001), Dahlby (Reference Dahlby2008), Kaplow (Reference Kaplow1996, Reference Kaplow2004), Jacobs et al. (Reference Jacobs, Mooij and Armstrong2009), and Jacobs (Reference Jacobs2018), among others. This literature does not assume homogeneous agents and a representative consumer, but heterogeneity in skills or preferences. Such heterogeneity is essential for justifying and understanding the benefits of redistribution.
2.3 Formal MEB and MCPF definitions
Briefly discussing some formal definitions of MEB and MCPF can clarify the basic theoretical concepts used above. Let us first briefly introduce some marginal excess burden (MEB) definitions. The excess burden (EB) or total deadweight loss of a tax system is the difference between the welfare losses caused by it and the tax revenues it generates. The MEB of taxation is the additional excess burden to raise an additional euro of tax revenue:
$$ \mathrm{MEB}=\frac{\mathrm{d} EB/\mathrm{d}t}{\mathrm{d}R/\mathrm{d}t} $$
From this general definition, MEB definitions for specific taxes in different settings can be derived. For example, one can show that the MEB of a marginal increase of a linear consumption tax (
$ \tau $
) is equal to (Dahlby, Reference Dahlby2008, Equation 2.35; Jacobs, Reference Jacobs2018, Equation 15):
$$ \mathrm{MEB}=-\frac{\tau }{1+\tau }{\varepsilon}^c $$
where 
$ {\varepsilon}^c $
is the compensated income elasticity of consumption demand with respect to the tax rate.
In contrast to MEB definitions, there is no agreed definition on how the MEB relates to the MCPF. Early works (Stiglitz and Dasgupta Reference Dahlby1971; Atkinson & Stern, Reference Atkinson and Stern1974) suggest that the optimal provision of public goods has to be lower in order to account for distortionary taxes. These works gave rise to the idea that for public projects a “
$ \mathrm{MCPF}=1+\mathrm{MEB}>1 $
” rule has to be applied. The rationale for these rules is elaborated in detail in Dahlby (Reference Dahlby2008, Chapter 2), such as:
$$ \mathrm{MCPF}=\left(1+\mathrm{MEB}\right)c $$
where 
$ c $
is conversion factor, which is larger than 1 if normal good is taxed, and MEB is for the compensating variation case as, e.g., in Equation (2).
However, MCPF > 1 rules are not necessarily applicable, even if distribution concerns are ignored. For example, Ballard and Fullerton (Reference Ballard and D.1992) argue that modified versions of the Samuelson rule
$$ \sum \mathrm{MRS}=\mathrm{MCPF} \ast \mathrm{MRT} $$
can in some cases result in a MCPF that is smaller than 1 if the tax system is nonoptimal. Recall that MRS are the marginal rates of substitution between a public and a private good, and MRT is the marginal rate of transformation between the public good and the reference private good.
Moreover, it is by now well established that the MCPF also needs to include the social benefits of redistribution, as governments may choose distortionary taxes that aim to increase social equality. For example, Jacobs (Reference Jacobs2018) defines the MCPF as “the ratio of the social marginal value of public income and Diamond (Reference Diamond1975)‘s measure of the social marginal value of private income”. From this general definition, MCPF definitions can be derived for individual taxes. For example, for the linear consumption tax, the MCPF can be written as (Jacobs, Reference Jacobs2018):
$$ \mathrm{MCPF}=\frac{1-\xi }{1-\mathrm{MEB}} $$
Note that income redistribution is valued if 
$ \xi >0 $
, which lowers the MCPF as compared to the case without social benefits of income redistribution (
$ \xi =0 $
). The tax is set optimally, if MCPF equals 1.
2.4 A general framework for benefit-cost analysis
In order to link the costs of taxation and the distributional benefits of taxation to welfare and the practice of BCA, a general framework for BCA is presented in Table 1.
Net benefits and the costs of taxation and the benefits of redistribution over different income groups (“income redistribution”).
The top part of the framework covers the costs and benefits without any correction for MEB. The benefits (B) consist of direct and indirect effects. The latter consist of wider economic benefits, some of which pertain to the labor market (L); they show the welfare effects of behavioral changes due to the policy measure (see Atkinson & Stern, Reference Atkinson and Stern1974).Footnote 20 For example, introduction of a childcare allowance will often lead to more labor supply, in particular because mothers will seek paid work or want to work more hours.Footnote 21 The indirect effects or wider economic benefits are commonly included as part of the benefits B.
The cost of a policy measure (C) is the amount of resources sacrificed by the government and private stakeholders to implement the policy measure (see Romijn & Renes, Reference Romijn and Renes2013). Examples are the public cost of building a road or bridge, or the public cost of an investment in education.Footnote 22
The benefits (B) minus the costs (C) result in the balancing item “net benefits” without any correction for MEB. This BCA-balancing item (S) is obtained by aggregating costs and benefits following the principle “1 euro is 1 euro,” irrespective of who gains or loses. In particular, no account is taken of whether the recipient is poor or rich. This principle is also known as the Hicks-Kaldor potential compensation criterion. If these net benefits are positive, the policy measure can be interpreted as a potential Pareto-welfare improvement: the beneficiaries of the policy measure could compensate the losers and still be left with a gain. This would result in a real Pareto-welfare improvement. However, in practice, those who lose from the policy are usually not compensated and it would also be very difficult to do so without any extra costs and without any additional behavioral changes.Footnote 23
The bottom part of the framework shows the possibility of adding corrections for the cost of taxation and benefits of redistribution of income. The cost of taxation (E) refers to welfare loss caused by distortionary taxation, i.e., the marginal excess burden of taxation (MEB). In addition, two types of distributional benefits are distinguished: first, distributional benefits of taxation (F) and, second, distributional benefits of the policy measure under investigation (M).
If net benefits based on the Hicks-Kaldor criterion (S) are corrected for the costs of taxation (E) and the distributional benefits (F and M), this results in a BCA-balancing item based on a comprehensive welfare measure (W).
The way the framework is presented above with only one column for costs and benefits (see Table 1) suggests that the primary purpose of BCA is to assess whether the benefits of a policy measure exceed or justify its costs. However, the primary purpose of BCA is to help select the best or most appropriate policy measure. This is also stressed by the federal BCA guidelines in the USA (OMB, 1992, Circular A94, general principles 5.c.3):
“Evaluation of Alternatives. Analyses should also consider alternative means of achieving program objectives by examining different program scales, different methods of provision, and different degrees of government involvement. For example, in evaluating a decision to acquire a capital asset, the analysis should generally consider: (i) doing nothing; (ii) direct purchase; (iii) upgrading, renovating, sharing, or converting existing government property; or (iv) leasing or contracting for services.”
As a consequence, different columns should be introduced showing the costs and benefits of a policy measure and its major alternatives, and whenever relevant, also comparing different ways of financing them.
Starting from this framework with different columns for different alternatives, the question whether a correction should be made for the marginal excess burden of taxation can be decomposed and translated into two questions. The first question is: Are the marginal distributional benefits of taxation (F) equal to the marginal distortionary effects of taxation? (E) If this is true, then no correction for the MEB is needed and the MCPF is equal to 1 (see Jacobs et al., Reference Jacobs, Mooij and Armstrong2009; Jacobs, Reference Jacobs2018). The second question is: How should the distributional effects of the policy measure for different income groupsFootnote 24 be accounted for? Does taking such distributional effects serious imply that the Hicks-Kaldor criterion should be abandoned and be replaced by distributional weighting of costs and benefits?
A proper response to these questions implies that different kinds of argument are taken into account: economic-theoretic, empirical, practical, and political. An economic-theoretic argument is that some effects could, for theoretical reasons, cancel out or could only be relevant under strict theoretical conditions (see Section 3). Arguing that some effects are relatively small or will be hard to quantify and translate into monetary terms reliably is an empirical argument (see Section 3). The cost and time of extra analysis is more a practical argument (see Sections 3 and 4). A political argument is the politicians’ wish to strictly separate issues of efficiency and equity (see Section 4).
3 Is the marginal cost of public funds equal to one?
3.1 Policy measures financed by general tax revenues
This section discusses whether the MCPF is 1, that is, whether a correction for the MEB is needed. We assume that the policy measures are financed by general national tax revenues, which is the case in which there is no clear relation between a policy measure and its way of financing. In the next subsection, the implications of other ways of financing are discussed. In the discussion, typically four arguments exist why no correction is needed:
(i) Distributionally neutral financing;
(ii) Optimal taxation;
(iii) Consistency with the current tax system;
(iv) Uncertainty about the size of MEB and distributional benefits of taxes;
The merits and limitations of these arguments are summarized in Table 2 and discussed in more detail subsequently.
Arguments in favor of assuming MCPF = 1 when the policy measure is financed by general tax revenues.
3.1.1 Argument 1: Distributionally neutral financing
According to Kaplow (Reference Kaplow1996, Reference Kaplow2004) the income tax can often be adjusted to offset the benefits of the public good.Footnote 25 In the case of a uniform monetary benefit to individuals, e.g., for a park or a bridge, a lump sum tax can be used. In case of benefits proportional to income or wealth, a proportional rise in the income tax can be used. More generally, distributionally neutral financing implies that those who benefit from a policy measure should also pay for it and that those who lose should be compensated. For example, in the Netherlands, the introduction of a totally new health care system in 2006 was accompanied by policy measures compensating negative income effects. Kaplow proposes that benefit-cost analysis of a policy measure should be split into two steps: first, an analysis of the policy measure financed distributionally neutral and, second, an analysis of a purely redistributive adjustment to the tax system.
However, distributionally neutral financing is only possible under very strict conditions.Footnote 26 These conditions will hardly ever be met for most policy measures and certainly not for policy measures financed by general tax revenues. So, the argument for distributionally neutral financing is usually not relevant for the practice of BCA.
3.1.2 Argument 2: Optimal taxation
In a world with distortionary taxes and redistribution, economic theory predicts that the costs of taxation are balanced against the benefits of redistributionFootnote 27 (see Jacobs et al., Reference Jacobs, Mooij and Armstrong2009; Jacobs, Reference Jacobs2018). If the current tax system is optimal, then the marginal costs of general taxation are equal to the marginal benefits of reduced income inequality. In terms of our BCA framework (see Table 1), F is equal to E and MCPF = 1. This implies that for welfare measurement, the BCA balancing item (S) need not to be adjusted for the cost of taxation or the distributional benefits of this taxation. This conclusion depends critically on the assumption that the tax system and policy is optimal. If this is true, the theoretical conditions for MCPF = 1 to hold are not very stringent.Footnote 28
The design of an optimal tax system depends critically on the aversion to inequality. A higher aversion to inequality implies larger distributional benefits and therefore also the acceptance of more distortionary taxes.
An optimal tax system assumes that sufficient information is available for a proper balancing of the distortionary costs of taxation and the distributional benefits and that this information is used in a consistent way to choose the mix of taxes and tax rates. Zoutman et al. (Reference Zoutman, Jacobs and Jongen2016) show that the welfare weights for different income groups that are implied by the Dutch tax systemFootnote 29 are in general higher for high-income groups than for lower income groups. This is what you would expect of an optimal tax system if there were general aversion to inequality. But Zoutman et al. (Reference Zoutman, Jacobs and Jongen2016) also note some anomalies that cannot be reconciled with an optimal tax system. In particular in the lower part of the income distribution, the welfare weights rise with income, instead of falling. This conclusion for the Dutch tax system is in line with the relatively low social welfare weights for the working poor found in studies on the tax system in other countries, see, e.g., Bourguignon and Spadaro (Reference Bourguignon and Amedeo2012).Footnote 30
Optimal tax policy assumes that redistribution by the government is efficient given a certain degree of aversion against inequality. In the Netherlands, many different types of policy measures are used for redistributive purposes, e.g., progressive income tax, minimum wage, housing subsidies, and tax deduction for interest from mortgages. For many reasons (e.g., political strategic reasons or information problems about the policy measures’ efficiency for redistribution), this current mix may not be optimal and efficient for any specification of aversion against inequality.
These arguments about the optimality of the tax system and tax policy in practice imply that the assumption of optimal taxation is probably too strong and therefore cannot be used for justifying MCPF = 1.
3.1.3 Argument 3: Consistency with the current tax system
A substantially weaker assumption than optimal taxation is consistency with the current tax system. The current tax systemFootnote 31 can be assumed to be broadly consistent with the current political and societal preferences. This tax system will roughly reflect the preferences and decision-making power of the successive governments and their constituents. The resulting tax systemFootnote 32 is a specific combination of distortionary taxes and distributional benefits. Financing public expenditure with less distortionary taxes would also have been possible, but apparently the distributional benefits of the actual choice of distortionary taxes provide sufficient compensation for the welfare loss due to these distortions. A correction for the MEB is then not necessary. BCAs can then proceed using MCPF = 1 resulting in BCA-outcomes that are consistent with the preferences for (in) equality as laid down in the current tax(-benefit) system and with other government policy.
Assuming consistency with the current tax system and hence using MCPF = 1 in BCAs is also the only politically neutral choice. Any alternative assumption (MCPF > 1 of MCPF < 1) implies that the BCA-analyst regards the current tax system as insufficiently redistributional (MCPF < 1) or as too redistributional (MCPF > 1), which can be regarded as a political statement.
Although it is reasonable to assume that the current tax system reflects current political preferences on the trade-off between equity and efficiency, such political preferences may change over time. Furthermore, political parties in the opposition or other groups outside the government may have a substantially different aversion against inequality. These arguments might suggest that with a forward-looking view or from the perspective of groups outside the government, the assumption that MCPF = 1 for BCAs of specific policy measures could well be arbitrary and misleading. However, such a suggestion is invalid, as becomes clear by applying a two-step argument. Suppose in the near future people with substantially different inequality aversion will come to power. Then they should first adjust the general tax system in line with their substantially different inequality aversion. In a second stage, specific policy measures can be evaluated on the basis of their costs and benefits, and then again the reasonable assumption can be made that for marginal changes the distortionary cost of taxation is equal to the distributional benefits of these taxes. As a consequence, whether the specific policy measure will be evaluated in view of the current or future tax system will not matter for its outcome.
3.1.4 Argument 4: Uncertainty about the size of the MEB and the distributional benefits
Estimates of the size of the MEB of the current tax system vary substantially and depend critically on the assumptions used. The European BCA-guidelines on transport infrastructure (Bickel et al., Reference Bickel, Friedrich, Burgess and Fagiani2006) recommend not making a correction for the marginal excess burden of taxation (MEB) because of the uncertainty in the estimates. For example, Kleven and Kreiner (Reference Kleven and Kreiner2003) illustrate for OECD countries that the size of the MEB depends on the way of financing (higher average tax rate or more progressive taxes), the inclusion of the participation decision, and whether also nonlabor income revenues such as allowances and social benefits are taken into account.
Also for the Netherlands, estimates of the MEB can differ substantially depending on the assumptions used (see Table 3). According to Jacobs (Reference Jacobs2015), the MEB of general tax and social security revenues in the Netherlands is about 0.50, i.e., a tax burden of 50 eurocent per extra euro of public expenditure. Some very reasonable alternative assumption can lead to substantially different estimates. For example, in the standard model, the MEB is 0.38, but if labor supply elasticity is lower or higher, the MEB changes from 0.25 to 0.51. If also the distortionary effects of employers’ social security contributions are taken into account the estimate of the Dutch, MEB more than doubles (1.07).
Estimates of MEB in the Netherlands: standard model and assumptions and some alternatives (Jacobs, Reference Jacobs2015).
The distributional benefits of the tax system depend on the degree of inequality aversion and the specific method used (see Van der Pol et al., Reference Van der Pol, Bos and Romijn2017). For some assumptions and methods, the MEB and the distributional benefits are of approximately equal size, but with other assumptions, the MCPF can be larger or smaller than one. This occurs particularly if the degree of inequality aversion is chosen in such a way that the current tax system does not come close to the degree of redistribution that is considered optimal for the chosen degree of inequality aversion. Such a choice for the inequality aversion implies that the current tax system does not reflect and is not consistent with current preferences.
However, the empirical evidence does not force one to choose such an inconsistent set of assumptions. The empirical evidence allows one to choose a consistent set of assumptions in which MCPF = 1. Therefore, empirical evidence does not help much when it comes to the size of the MCPF. We are essentially left with the earlier argument about consistency.
According to Mishan and Quah (Reference Mishan and Quah2007, see, in particular, footnote 1 on p. 24, appendix 4 and 13), including a correction for the marginal excess burden of taxation is a common mistake in many BCA-textbooks. Their argument is not only that the estimate of MEB is uncertain, but also that it will in general be very small. Mishan and Quah argue that in the real world, there will be many different types of deviations from a perfectly competitive economy, e.g., due to monopolies, information problems, transaction costs, efficiency wages, regulations, taxes, subsidies, and external effects. In such a world, the net effect of a specific policy measure on overall distortions can be either positive or negative. As it concerns a policy measure of marginal importance to the whole economy, it will not have any significant effect on the distortions in the economy. The measurement of the net effect on welfare will be illusory and should therefore best be ignored.Footnote 33
Our conclusion is therefore that the argument of consistency with the current tax system (Argument 3) offers convincing ground for assuming MCPF is equal to one. The argument of uncertainty of the estimates of MEB and distributional benefits (Argument 4) is less convincing, but makes clear that empirical evidence does not help much in assessing the value of MCPF. The two other arguments, about distributionally neutral financing and optimal taxation, are purely theoretic arguments and do not help much in settling the debate on the best solution for BCA practice.
3.2 Alternative sources of financing
Policy measures can be financed from other sources than general tax revenues. Examples are toll fees and congestion charges, social security contributions, public-private partnerships, local taxes, loans,Footnote 34 and a mix of financing by central and local government. Would our analysis about MCPF = 1 then still apply?
All these other types of financing should be regarded as a separate policy measure that should be analyzed separately in a BCA. It would be an analysis similar to a change in a specific tax or a general revision of the tax system. For all such policy measures, the financing issue of the policy measure can best be ignored. As a consequence, the question whether MCPF = 1 is not relevant, only measuring the costs and benefits, including the distortionary effects of the policy measure as such (L) and the distributional benefits of the policy measure (M).
Different policy measures may have positive and negative synergies. An example is the construction of a new road financed by (an increase in) a congestion charge. This case should be analyzed by a BCA of the construction of the new road financed by general tax and a separate BCA of the introduction of a congestion charge.Footnote 35 Finally, both measures can be combined in a joint BCA. This combination may in some cases reveal that broadly those who benefit should also pay or that the financing is distributionally neutral (Kaplow’s argument, see Section 3.1).
Financing a road via a congestion charge may also be compared in a BCA with various other ways of financing, e.g. (an increase in) car registration tax, (an increase in) excise duty on petrol or a public-private partnership. In each of these cases, the financing measure is a separate issue and should be analyzed separately, while taking into account their distortionary cost and their distributional benefits.
Another case in point is extending the basic health care package of the Dutch government with a new long cancer medicine. This can be financed by raising general health care social insurance contributions, but also by income dependent social insurance contributions, lifestyle-dependent contributions or by out of pocket payments. The merits and limitations of such an extension of the basic health care package should preferably be analyzed for different types of financing. This logic also applies to major policy changes accommodated with a package of compensating policy measures. They should be first analyzed separately and then jointly.
The way policy measures are financed is often also a strongly politically motivated choice. Who should pay and to what extent is a normative issue in which political criteria like justice and solidarity are important. To what extent a BCA can be helpful in such issues of equity will also be a topic in the next section.
4 Distributional benefits of the policy measure
Nearly a century ago, Pigou emphasized the general principle that for optimal public expenditure the social gain from a marginal increase in resource use should be the same everywhere. This principle ensures not only an efficient balance between resource use in public and private sectors (see Sections 2 and 3), but also between resources used in different parts of the public sector. He illustrates the latter with the following example:
“Expenditure should be distributed between battleships and Poor Relief in such wise that the last shilling devoted to each of them yields the same return of satisfaction” (see Pigou, Reference Pigou1928; Sandmo, Reference Sandmo2011, p. 261).
This quote indicates that distributional benefits of a policy measure should certainly not be ignored in a BCA.
However, the net benefits in a BCA following the Hicks-Kaldor criterion (S), give an equal weight to the effects of a policy measure for low and high incomes (1 euro is 1 euro irrespective of who gains or loses; see Section 2). As a consequence, distributional benefits (or costs) from a transfer from high income to low income (or vice versa) are ignored. A choice for MCPF = 1 implies that the distributional benefits of taxation are included. Consistency then suggests that also the distributional effects of the policy measure should be accounted for.
Two solutions are possible to remedy this. The first is to quantify the effects on various income groups and show these effects separately from the BCA balancing item, e.g., to present them as a token entry (p.m.). This would be on a par with the treatment in BCAs of any other effects, e.g., effects on biodiversity, that are mentioned but not translated into monetary (welfare) terms and therefore are excluded from the BCAs net benefits.
The second solution is to incorporate the distributional benefits in the BCAs net benefits by using distributional weighting, i.e., to give a higher weight to benefits for low-income groups than for high-income groups. In this way the distributional benefits (M) are included in the BCA balancing item, possibly as a correction to net benefits following the Hicks-Kaldor criterion (see, e.g., Harberger, Reference Harberger1978; Mishan & Quah, Reference Mishan and Quah2007; Florio, Reference Florio2014; Hendren, Reference Hendren2014; Zoutman et al., Reference Zoutman, Jacobs and Jongen2016).
The first solution requires data and models that allow the calculation of effects for different income groups. These are in principle available or could be developed, but require a lot of additional analysis. The second solution requires taking another step, which primarily involves establishing a set of weights for different income groups. van der Pol et al. (Reference Van der Pol, Bos and Romijn2017) show that various methods are possible and that for each method also various more operational choices are to be made.
In the Dutch BCA working group on the MCPF, the merits and limitations of distributional weighting were extensively discussed (see Table 4). The conclusion was that, though the Hicks-Kaldor criterion is not politically neutral, it is nevertheless a useful benchmark for political debate, in particular when also major effects for various groups of income are shown when relevant. The more ambitious solution of incorporating distributional benefits in the BCA by also adjusting the BCA balancing item using distributional weighting was rejected. This was considered to have major adverse effects on the use of BCA in general in Dutch political decision-making.
Should benefits of redistribution over income groups be included in the BCA balancing item?
This conclusion is in line with international BCA practice. In international BCA practice welfare weights, i.e., different weights for different income groups, is hardly applied. This is even true when their BCA guidelines recommend it. Major cases in point are BCA at the World Bank and in the UK. Since the 1970s, BCAs are a major trademark of the World Bank. According to the old BCA Guidelines of the World Bank (Little & Mirrlees, Reference Little and James1974) benefits for households in BCA should take account of their income, in particular when a policy measure is aimed at improving the situation of the poor. However, in the World Bank’s BCA practice (see World Bank, 2010; Van ’t Riet, Reference Van ’t Riet2016) this was hardly ever done. One reason is that it would have made BCA for projects in development countries even more difficult. A second reason is that BCAs were often not relevant for financing a project. A third reason is that BCAs were mostly used for sectors in which redistribution of income was considered to be a secondary issue, e.g., for agriculture, energy, transport, and water. For policy measures with respect to education, health care, and the environment, hardly ever a BCA was made.
The BCA guidelines of the UK Ministry of finance (“UK Green Book”, Treasury, Reference Treasury2011) support the idea that in assessing cost and benefits also the income and wealth of households that gain or lose should be taken into account. However, it is also remarked that compiling information on this would often lead to disproportionally extra costs. Furthermore, no explanation is given for what type of policy measures or under what kind of circumstances providing such additional information is needed.
5 Conclusions
According to economic theory, taxation drives a wedge between private and public benefits, which distorts labor supply, consumption, and investment and leads to loss of welfare. One would therefore expect that in BCA of public expenditure a correction is made for the costs of taxation, i.e., for the MEB. However, looking at BCA practice all over the world, textbooks on BCA, various specific BCA guidelines and economic literature, no consensus exists about such a correction. In most countries in the world, no correction for the MEB is made, in many BCA guidelines the issue is not even mentioned, while BCA textbooks and economic literature disagree on the need for a correction.
This paper provides an overview of the theoretical, empirical, and practical arguments in favor or against an MEB correction. It argues that a pragmatic approach for BCAs is to assume firstly that a policy measure is financed via general tax revenues and, secondly, that the MEB is broadly counterbalanced by the benefits of redistribution of these taxes. The latter assumption is consistent with the preferences for equality in a country’s current tax system and is a simple, pragmatic, and politically neutral assumption. This assumption does not imply that the tax system is optimal or that BCAs should be distributionally weighted. This pragmatic approach implies that the MCPF is equal to 1 and then no correction is needed in BCAs for the MEB.
A policy measure may also be financed by an alternative source of financing than general tax revenues, such as a road financed via congestion charges or by higher excise duties on petrol. These alternative ways of financing should be regarded as separate policy measures that should be analyzed and compared separately in a BCA; this comparison includes also their distortionary costs and their distributional benefits.
